Support the show to get full episodes and join the Discord community.
Mac and I discuss his systems level approach to understanding brains, and his theoretical work suggesting important roles for the thalamus, basal ganglia, and cerebellum, shifting the dynamical landscape of brain function within varying behavioral contexts. We also discuss his recent interest in the ascending arousal system and neuromodulators. Mac thinks the neocortex has been the sole focus of too much neuroscience research, and that the subcortical brain regions and circuits have a much larger role underlying our intelligence.
- Shine Lab
- Twitter: @jmacshine
- Related papers
Mac 00:00:04 I was like, oh my God, around the corner in this place that everyone likes to forget. And if you tell most neuroscientists about it, that aren’t studying it, they’ll yawn and pretend like they’ve got some other problem to solve, but come on, man. The cerebellum is just beautiful in a way what the thalamus is doing is really controlling the state and then any influence that happens to it or the time, whether it be a cortico influence, a basal ganglia influence, cerebella a curricular influence, a neuromodulatory influence is going to shape and change the way that the state will change over time, which is one of the most crucial factors for determining how we do what we do. These things often strike you when you least expect them. But I think they’re an underappreciated aspect of, of science or at least the pilot sites that I really love of that kind of wallowing in your uncertainty until it resolves itself, I think is one of my favorite pots.
Speaker 0 00:01:00 This is brain inspired.
Paul 00:01:13 Hey everyone, it’s Paul. On this episode, I bring you an appreciation for the detailed nitty gritty work being done in systems neurobiology that highlights its importance in understanding the big picture functioning of our brains. Mac shine runs the shine lab at university of Sydney in Australia, focused largely on how systems neurobiology can help us understand our cognition. We talk about a pretty wide range of topics, all of which dance around systems neurobiology, which is on the whole, what Mac focuses on, but that is a vast range of topics. One of the main things we discuss is the role of sub-cortical brain areas that don’t get nearly as much attention as the neocortex gets, especially in the neuro AI world, where AI tries to glean some inspiration from brains, but work-like max theoretical work that we discuss hopefully will change that cortico centric bias. Many of us have.
Paul 00:02:10 The main thing we discuss is the role of the thalamus mediating communication among the basal ganglia, the cerebellum and the cortex. And the idea is that through these interactions, the thalamus serves to nudge the brain into different dynamical states of operation, based on the ongoing demands of the organism and the context of the environment. And Mack has laid out theoretical roles for the basal ganglia and the cerebellum that don’t necessarily align with the way we traditionally think of them, nor does the thalamus story align with its traditional story. But Mack is interested in the system as a whole and also studies the ascending arousal system and the neuromodulators, it deploys to affect the state of our cognition. So we discussed that and how he believes it’s helpful to think of all this complexity from a lower dynamical systems perspective. So this is a heavy systems neurobiology episode, but well, worth your time.
Paul 00:03:07 And I think you’ll find it worth revisiting. If you’re interested in forming a zoomed out broad picture of how complex systems like brains work on that note, I want to add that Mack. And for that matter, plenty of other guests I’ve had on the podcast to me is an example for all of this, whether you’re an aspiring student or beyond, because he’s an example of someone who makes it clear that given enough focus over time and earnest interest in your questions and a persistent curiosity in the face of resistance, it’s possible to get to a point where your thought becomes more fluid and facile and able to navigate among different systems and concepts and form some appreciation for the whole. And that is something to behold and something I’ve always struggled with in my own pursuits. I think Mac would tell you it never becomes easy, but I also think listening to him, you may disagree and think that at least it becomes much easier. You can learn more about Mac in the work that we talk about in the show notes at brain inspired.co/podcast/ 121. Thanks for listening. I hope you enjoy Mack, including his dad joke
Paul 00:04:19 6:00 AM, 6:00 AM. What are you, what are you doing up at 6:00 AM?
Mac 00:04:25 Well, you know, uh, when you live in such a beautiful part of the world, waking up early, looking out over the, you know, beautiful green, uh, you know, outlooks over the house, how could you not take advantage?
Paul 00:04:36 What time do you actually wake up normally?
Mac 00:04:38 Well, uh, that really depends, uh, on my kids, Paul, um, I have fairly rambunctious boys that like to jump up at the crack of Dawn and get into all manner of hi-jinks I’m my wife and I are often up quite early. And,
Paul 00:04:51 Uh,
Mac 00:04:54 Yes, yes. Uh, and then, you know, take, take advantage of the day. If the, if they don’t wake up, then I get to have a nice coffee in silence and do a bit of reading.
Paul 00:05:04 I’ll see mine sleep in a little bit now until around seven. And, uh, I get up around four 30 or five just because that is the only real uninterrupted time that I get. Right. So, uh, so I was just curious if you were in that same boat,
Mac 00:05:19 Uh, yeah, on a good day. I get to get up and have, have, you know, read a local papers or something like that, but it seems like more and more, uh, I get to embrace my role as a father of young, uh, young, fun boys at that time.
Paul 00:05:33 How old are they? We don’t have to talk about this for everybody.
Mac 00:05:35 How old are they? No, no, I’d love to, uh, so Tyler’s 10 and Calen seven. Um, they’re, they’re riding that fun age of, you know, sport and video games. So
Paul 00:05:45 Yeah, I was my son’s soccer coach this last season, which was, um, challenging and also fun of course. Okay.
Mac 00:05:53 Well,
Paul 00:05:54 Oh, you did well, like a football soccer. I don’t, what do you guys call it down there?
Mac 00:05:58 Well, we call it soccer, but yeah, the Europeans would call it football and we had a kind of married band of pranksters and his team was a lot of different kids from a lot of walks of life and different skill levels, but we had a lot of fun.
Paul 00:06:12 Yeah. All right. Well, good morning, Mack. And thanks for joining me here on the podcast. Um, it’s been awhile actually. I’ve been meaning to have you on for a while. And, uh, you’re on the, I keep a list, uh, from my patron supporters of, uh, suggested guests and you’ve been on that list for a while. So I’m glad to have you on here.
Mac 00:06:30 That’s awesome.
Paul 00:06:32 The first thing I want to ask you, do I understand this correctly because, well, we’ll get into what you do here, but do you have a medical background? Did you go to medical school?
Mac 00:06:42 Yeah, that’s right. So, um, I did a undergraduate in kind of a combination of like biochemistry and psychology and really liked a lot of parts of it, but other pots really didn’t resonate. Um, and then went to medical school at the university of Sydney. Um, loved parts of that. Um, really the thing that just absolutely resonated with me was this challenge of trying to kind of take in all this information from all these different systems, the cardiovascular system, the spiritual system, the liver, the kidney, and understand this sort of cellular level physiological detail, not just the anatomy, but how it actually worked, but then how it broke down as well in different disorders. And so you were basically this rapid fire cycling through these different, uh, um, sort of systems and their pathology, but then there’s this really amazing point where for me, it was somewhere in about the second year of medical school where you sort of, it’s like when you’re wandering around a foreign city that you’ve been through maybe one or two times before, and you sort of like, think you’re completely lost and you turn the corner. You’re like, oh wait, that’s where my hotel is. And you like, figure out how the bits all kind of intersect. And that I think I became addicted to that, that idea of deep diving into problems and then finding a way out of the morass of uncertainty into that solution space. And so I, you know, that’s really stayed with me, I think, uh, ever since I was in medical school.
Paul 00:08:11 Well, yeah. I want to ask you more about this. I mean, I was, what I was wondering is if, if that’s, if ma if your medical background is why you seem to at least love the anatomy so much, because it always gives me the spins, see you, you saw it as a, uh, a challenge to integrate these systems. And I just ran in fear essentially.
Mac 00:08:30 Yeah. Maybe. Um, I think I had it beaten into me in the early days or something. Um, no, look, I think, you know, at the end of the day, you know, anatomy is kind of this, uh, this ultimate arbiter of you can have the most beautiful idea in the world, but it’s like that little meme that sort of data pops up and then shakes its head and you’re like, oh, well, you know, the, the answer doesn’t really fit. The, the anatomy kind of just is what it is it’s sitting there. You can see, and there’s so much beautiful work in the field now just incredibly detailed molecular analysis of the, of the brain that if you’re up in this sort of space. So trying to figure out how this bit interacts with that bit, if you can kind of try to lay some foundation under it with anatomy that kind of has those details baked in.
Mac 00:09:15 I think that really stood the test of time in medicine and physiology. You know, if we understand the function of the hot, via understanding the particular types of cells, the sinoatrial node and the ventricular node and the way the muscles contract in particular coordinator way, then we can kind of like lean our theories on that. We can understand how the Frank Starling law of contraction relates to that physiology and we can kind of build up from there. And so for me, I really do find myself a lot of time thinking of ideas and thinking of kind of implications of anatomy, but then kind of come back to work out, okay, how could that fit with the other things we know? And there are lots of times when it can be a really frustrating endeavor. Um, the anatomy of literature is, uh, is amazing, but it’s not perfect in any way, shape or form.
Mac 00:10:06 And one just to give one really good example. So much of what we know about anatomy comes from model organisms. And we don’t really know whether or not exactly the same anatomy exists in humans. And a lot of the times it’s really different in really funky ways that we don’t quite understand the implications for. And so you end up having to be very cautious about reading a particular literature at the GearCage, if it were the same in humans, what implications might that have for some psychological function I’m interested in, but you can’t really take that to the bank all the time. So you’ve really got to take a lot of these things with a grain of salt and just use them as a sort of two-way conversation.
Paul 00:10:41 Yeah. I mean, one of the things that is really impressive about your work, um, is it has a holistic feel, right? And I mean, this has, because of the struggle you were just talking about. Um, but you come at it with, from so many different angles and seem to integrate so many different things. You’ve talked about this a little bit already, but, uh, you know, people have different opinions on the right approach to take, to study, uh, intelligence in general, right? So there’s a, a loud, powerful cry these days to take a top-down approach and think about the behaviors, understand the behaviors, think about the computational level of Mar and then use that to then just simply look for it in the brain and, uh, and confirm it right on the other hand, uh, there are people like Steve Grossberg, who, whom I’ve had on the show who, um, doesn’t think about, well, traditionally didn’t think about the brain at all. He sat with these psychological data and then built neural networks thinking about how those data could be explained right. And implemented. So how would you describe your approach? Do you always start with the anatomy, do you, or is it just a, a messy, uh, cycle that, well, just let you, I’ll just let you answer how about,
Mac 00:11:56 Yeah. I’m afraid this is one of those situations where, um, you know, getting to meet the chef, you’re like really that’s what goes into this. Um, so yeah. Um, I think, you know, maybe another kind of missing pinnacle in, in that, uh, that description as well would be the kind of brain from inside out approach, uh, URI Zaki where you would say something like once we understand the degrees of freedom inherent within the nervous system, then we could try to launch from there to try to work out how they mesh onto the kinds of things that we can clearly do the kinds of, sort of functional capacities we have. And, and, you know, I, I’m a little bit of a kind of perspective list. Um, and I think that different questions lend themselves to a starting point from the different ends, but, but, you know, because I don’t really have skin in the game and I’m not having to advocate one of the positions that I think is maybe underrepresented.
Mac 00:12:53 I probably, you know, will just sit on the fence and say, we need to kind of make them both talk together, all three of them talk together. Um, and so when I think about Moz framework, we often think about right, we think about the computational at the top, the kind of problem that’s trying to be solved. And we think about the implementation that, you know, the way it’s actually kind of baked into the, the animal. So we think about flying, uh, with a bird. And then we think about the feathers and the wings and the muscles in terms of the limitation. Um, the algorithmic level is often one that people think of as this kind of special level of doing something different. But for me, how I think about it is the algorithm is algorithmic level is where the computation meets the implementation. It’s, it’s how the feathers and the, and the wings interacted to give rise to flight.
Mac 00:13:37 There’ll be in terms of a plane, how the shape of the wings and the velocity of the plane allows the system to take off and fly. Um, but to me that that’s a different kind than the computation and the implementation. It’s, it’s almost like a kind of transfer function between the two of them. And so I think if, when we say we’re interested in the algorithmic level, I think we’re kind of almost competing to this notion that we agree that the implementation and the computation of both matter. And it’s about how, if we care about that particular problem, let’s say how a brain solves a working memory challenge or how we remember a particular phenomenon, or maybe even how a convolutional neural net is able to classify between a dog and a cat, whatever it is that we care about. If we care about the Albert algorithmic level, we’re saying something about the architecture and something about the problem mattered, and it’s how those bits came together in this particular step that I, that I want to understand.
Paul 00:14:28 So, uh, do you think in terms of constraints, um, and so I know you, you are a fan of the dynamical systems theory approach, and you use it and a lot of your work, um, but do you see both the implementation level and the computational level as providing the, I mean, what, you know, is constraint and important constraint to, uh, from, from both sides to then mediate the algorithmic level?
Mac 00:14:54 Yeah, it’s a, it’s a really great question and pull. And, um, this is where I think the, the chef in the kitchen out of analogies comes to fruition because at the end of the day, I think with my own thinking, when I reflect back on how I came to a particular, uh, sort of hypothesis for some interaction than other system, it’s usually the ignorance that I have that allows me to get to that position. And, and, you know, my, my, my Paul poor understanding of the specifics of anatomy or my, my, uh, less than subtle appreciation for the complexities of some cognitive capacity that we have allows me to kind of mush them together in a way that says, well, what if they are interacting in this way? And if they are, what are the implications? And I think for me, that’s really the thing I love the most about science is the kind of hypothetical nature of it.
Mac 00:15:46 Right. I, my job as a scientist, as I see it is to say, um, I don’t know if this is right, but if it is, what are the implications, and then I go off and try and test some of those. And sometimes the ideas of good and other times the ideas are really bad. And I think that I’ll appoint of a scientist isn’t to sort of, sort of say, hold up a tablet, etched in stone, here are the correct things. Here are the facts of the world. But rather to say, someone who is trained in the scientific process knows how to go out and say, that’s curious, how the heck does that work? How can I put those things together? And then it bring it back to the kind of anatomy thing. This is like a fundamental mystery to me, somehow in some way, these tiny little specialized cells that just really, really love to bother each other all the time with action potentials or to squirt neurochemicals onto one another, that changed the, you know, the way that the different systems can fire action potentials each other.
Mac 00:16:38 Somehow that coordinate activity gives rise to this conversation. And what a fun challenge, how do we work that out? And what is the language we need? What are the constraints that we need in order to take this bag of tissue? I mean, this is again where medicine kind of helps, right? Because early on in my, in my learning, you’re, you’re, you’re there in the wet lab, looking at cadavers and, and pro sections and realizing that the system is fiscal like deeply physical at the end of the day. And it can, it conforms to all of the same rules and laws that have given rise to really great explanations of the heart and the liver and the kidney, right? It’s the same sodium potassium ATPase and all of those cells that shoveling ions across membranes, it’s the same into a plasmic reticulum. It’s storing calcium that then get real, it gets released when you need it to force some change in the action of the system, right?
Mac 00:17:30 That’s the, that’s the mechanism of your heart increasing. When you go for a run, it’s the same thing that your brain uses when you’re trying to increase the firing rate of neurons as a function of something like noradrenaline or dopamine or acetylcholine, some of these, these gain mediated mechanisms. So to me, there’s a, there’s a, I have a deep reverence for the fact that we have this huge challenge on our hands. And as a scientist, I want to be able to come in and say, well, I don’t know how it works, but what if it worked like this and the anatomy kind of doesn’t fit with that story over there, but it does suggest this one and just sort of help put constraints into that framework. So we can start asking empirical questions and, you know, this is, you know, this is where I sort of see neurosciences at a really early stage and a really exciting one because literally these data we have and look at this really hard problem we have, and we can kind of attack it together.
Paul 00:18:19 You want to talk some science?
Mac 00:18:22 Of course.
Paul 00:18:24 Uh, another reason why I was interested reading your work is because, so we’re, we’re about to talk about the role of the thalamus and, uh, sub-cortical structures like the basal ganglia and the cerebellum. So these days the cortex is the most important thing in the brain, right? And it has been for awhile. And at least the, uh, if you ask, uh, I don’t know, eight out of 10 neuroscientists, something like that. But, um, when I was going into graduate school, I was offered a project from my advisor mark summer. So we studied a cortical area called frontal light in front of my field, uh, which has loops, uh, that travel through the thalamus and basal ganglia and loops that traveled through the thalamus and the cerebellum. And he wanted me to, well, he offered me a project, uh, for the cerebellum and the frontal eye field thalamus cerebellum loops.
Paul 00:19:12 Um, I ended up doing my own silly project on metacognition, but then reading your work, I kept thinking what if I had done that and that, and how involved I would be, uh, with and familiar with, with this sort of stuff that you’ve done. So, I don’t know, I just, uh, a little little trip down memory lane, partially reading, reading your work. So, so like I said, it’s all cortex, right. But, um, one of the things that you have done is, uh, brought in the thalamus and, um, and loops in the thalamus with sub-cortical structures. So I want, uh, I would love for you to just kind of summarize why you think the thalamus is important. And then, uh, just the overall broad picture of the cerebellum and the basal ganglia, and you can get into whatever nitty-gritty detail that you would like. And of course, I assume that you’re gonna talk about dynamical systems theory as well, and I’ll interrupt you.
Mac 00:20:02 Yeah. And do it all in the auditory format to where I can’t show any pictures.
Paul 00:20:06 And by the way behind you, is that are those, uh, Ramonica hall drawings on the wall?
Mac 00:20:11 Well, it’s a, it’s a Greg Dunn picture, but it’s, I think it’s inspired by the golgi statins from keyhole of the left five-prime neurons, which, uh, I, I have to say Matthew lock-in in Berlin has convinced me, these are the powerhouse of the cortex. They’re just beautiful, beautiful cells. Maybe we’ll talk about them in a little bit. Yeah. Um, so yeah. Um, you know, I think a historical perspective helps a little bit to orient this and then maybe we can get into some of the details. So yeah, the cortico centric perspective, um, has really, I think, defined our epoch of, of neuroscience and, and there’s really good reason for it. If you were trying to understand the human brain and you look at it, right, you just put it on addition in front of you. There’s a whole bunch of cortex there. And if you compare it to a chimpanzee or a macaque is a bunch of more of the cortex right in front of you.
Mac 00:21:02 It’s a really great place to start looking just from first principles. And also when you see people that come into the clinic with, let’s say a particular stroke, this idea of localization of function in the cortex is really, really profound, really pervasive. And it’s been there from, you know, for a really long time. In fact, it goes all the way back to the Coliseum. Galen was the physician for the, uh, for the Coliseum for a few years. And he actually sort of came up with this hypothesis that for him nervous system function was actually about pneumatics. It was about pushing fluid through holes and tubes and things. But for him, he was like, okay, if the gladiator comes in, you know, and he’s had the kind of broad sword hit him across the brow. Now, all of a sudden he can’t, you know, um, you know, uh, make any mental plans kind of like, uh, gauge, maybe couldn’t inhibit his own behavior.
Mac 00:21:52 And then you’ve got someone over here who received a mace to the occipital cortex where he can’t see anything on the right. And so he came up with this notion of the Numa flowing from these very particular locations. And that has really kind of spread over time to this. I think this sort of, I would call this kind of localization kind of hypothesis is, is really the dominant story in the field. And, and really like my, the world that I came from, uh, so empirically it was using functional MRI to run analysis of brain imaging data. And Don’t apologize. We can w we can talk about Fri, uh, in a little bit, if you want. I mean, in a lot of ways I think ephemera is, um, is kind of like a, like a pimply teenager. That’s like been through it. All right. They’ve been insulted in every way they possibly could. So like, bring it on, man. I can take it right. They’re kind of coming out onto their own,
Paul 00:22:44 Uh, favorite, uh neurophysiologist um, uh, uh, poking, poking someone who does have MRI. I have a lot of respect for that from our eyes. So,
Mac 00:22:52 No, sorry. No, no, no, no, no, no, not at all. But so the, the prevailing, uh, you know, way of analyzing it from right data, which really kind of came from pet imaging, which was, is to run an experiment, right. Um, run a block of looking at a face and a house and then use some statistical contrast to find the most extreme statistical values. Yeah. And that, that approach really lends itself to this localization idea. Right? If I get you to look at a picture of someone you love and someone you don’t like very much, and I control some, you go, I’ll look love is sitting in the nucleus accumbens. That’s a really sort of attractive conclusion for me to make. Um, but if you think about it, uh, you know, just from a, again, from a slightly different perspective, if I was to sort of take that part of your brain and, you know, somehow expertly remove it and put it onto a dish, it’s not loving anything.
Mac 00:23:38 It’s just a bit of neuron so that those neurons have to interact with the whole to be able to function. And so we need to start thinking, I would argue at a much broader level about coordination amongst nervous system, uh, areas. We still need to keep a skerrick of that localization idea. Cause it’s clearly got some truth to it. There is a specificity to our nervous system. This part of the brain is different to that part of the brain. But if we take that to its extreme, I think we, we come up with so philosophical answers that are a bit impoverished. If we start to think about how that individual area with its constraints works with the whole system, admittedly, a very hard problem. We can start to ask about how the system can actually use the inflammation that might be being processed in that area for adaptive function.
Mac 00:24:21 So once you start moving in that direction, you can start thinking about different parts of the cortex interacting with one another. You could say, how does the frontal cortex interact with the pridal cortex during working memory or something? Or you could ask, how do I, you know, if we wanted to talk about something like pretty deep processing, how is some higher level area providing some kind of a prior, some kind of evidence or information for that system so that when I press it processes evidence, it can push it in one direction and interpret them one way or another. Um, those are one style of questions you can ask them. The field is really, you know, uh, is a really developed field in that space. But for me coming from that background again in medicine where you can’t think of the hot outside of the context of the whole cardiovascular system and the lungs, I’m forced to kind of come back to that perspective of saying, what are we leaving out of the picture?
Mac 00:25:13 What if we look at just the cortex, what else is there? And if you start to zoom out just a little bit like that, you notice that number one, the cortex does nothing on its own, right? It’s interact, it’s interacts with so much of the rest of the nervous system. Um, and to the consequences of impairments in these other areas, the thalamus is a really great example, but it’s by no means the only one, um, is really profound. And so if you have a stroke in your thalamus or a tumor in your thalamus, you often lose consciousness, you go into a coma or you can have really, really profound deficits in wide ranges of domains. They’re not as specific as you might see in the cortex, but they’re really, really profound. And so we’ve known for a really long time that the thalamus is incredibly important for all these really large functions. Arousal’s one of them, but it’s involved in working memory. It’s involved in attention, it’s involved in a ton of different psychological,
Paul 00:26:07 But the traditional story, right, is that it is important because it’s a pass through maybe a bottleneck for feeding the important stuff. Right. That’s the traditional story. And that’s the story that’s changing.
Mac 00:26:21 Yeah. But nobody puts Elvis in the corner. Right? Paul?
Paul 00:26:25 No. Well, not anymore. Thanks to people like you.
Mac 00:26:29 Well, so I think, you know, I, uh, I’m absolutely just the, um, the broadcaster of other people’s brilliant work. Um, so the wellness, yes, for a really long time, got put in this, in this basket, as you know, the relay was kind of, um, the story that people would tell him the reason for that is that if you look at the connections of the thalamus, let’s say in the context of vision, you’ll see the inputs coming to the retina, pass through an area called the lateral geniculate nucleus, which then goes on to the visual cortex as well, the superior colliculus and you’ll say, okay, well we know V one’s really important. And if I knock out V1, someone can’t see anything. And so what I’m just going to say is, well, the pharmacist’s job is to kind of make sure that the cortex got the information that it needed to process that vision.
Mac 00:27:14 And this again is sort of embedded within this. Um, uh, I would argue kind of a misguided view of, of evolution, which is that, um, what’s happened over time, is that a brainstem of the reptilian brain has kind of like had an ice cream put on top of it, the, the, the subcortex, and then this got like a sort of rain coat put over the top of it. And the cortex, this is called the McLuhan’s triune brain theory that essentially adding whole new pits on top of one another. Um, whereas a much better description of, of evolution. If you talk to someone like Paul CISAC or Lupo Ellis, is that you’ve got the basic bowel plan of the whole time. There’s, uh, uh, a midbrain, a brainstem, it connect to a spinal cord, a cerebellum, a thalamus, a tectum and some form and a hypothalamus.
Mac 00:27:57 And the some form of what we call a tellin CapitalOne, which is cortex special ganglia, a bunch of other structures, like the amygdala that have sort of expanded out like someone blowing up a balloon and the balloons degree of freedom is up in the telencephalon it’s expanded, but you’re not adding things onto one another. Um, and so the reason that that’s important is that if you think of the cortex as being added on, then it’s really, really easy to think. Well, in humans, the reason we can see the reason we’re conscious is that the cortex just popped in. And before that these poor suckers that didn’t have a cortex, they couldn’t do anything. They were just little automatons running around the world. Um, if you take the, um, so that was the kind of prevailing way of thinking about it. If you start from there, you can get some really, really interesting taxonomy.
Mac 00:28:37 For example, Mary Shannon, uh, regularly have a really nice way to think about the thalamus, which is that a lot of the connections are like that one in the LG. And I talked about what they’d call first order, which is that they receive an input and pass onto the cortex, but a whole bunch of the thalamus is actually what they call higher order, which means that it receives cortical inputs and sends it right back to the cortex. So it’s a lot like a kind of complicated, hidden layer in the kind of deep, deep learning kind of a way of thinking. And that it’s doing some kind of weird OIC, mentation or manipulation of the data, but it’s not directly related to what’s inputting or what’s outputting from the network. Um, so, so that’s one way of thinking about it, but the way that I’m really attracted to comes from a different scientist, Ted Jones is a neuroanatomist from, from New Zealand, um, unfortunately deceased.
Mac 00:29:20 And his way of thinking about it was a bit more thalamus centric. What he said was if we look at the thalamus and we try to understand the projections of the different thalamic nuclei, what kinds of patterns do we see? And he identified the type that we talked about for that relay type. He called that a core type where it receives an input, and it protects really precisely to the cortex, usually in the middle layers of the cortex. But interdigitated with that, it’s almost like a blend of these little bit ying and yang where every single nuclear and that the elements has a little bit of each of these different populations. Some of my, more than others, this other population, he called the matrix population. Cause it looked like the core nuclei were embedded in this matrix of these other cells in contrast to the corals, they protect up really diffusely.
Mac 00:30:01 And they do that either to the super granular layers of the cortex where a lot of the feedback projections from highlighters come in, um, but also to a lot of important sub-cortical structures like the striatum and the amygdala. So if we, so if we, if we start with this sort of Philando centric perspective, all of it’s just a relay then if we start incorporating these other cell types, which are really numerous, um, it’s really hard to take that same analogy of relay and kind of like feed it onto these other cells, right? If they’re relaying, they’re doing it in kind of a weird way, right? Th they’re sending an infant a message out in a really broad fashion, you know, things could get lost. You could have that, you know, purple monkey dishwasher game that we used to play or play when we were kids where you kind of, you can’t really envisage that, um, that message passing metaphor, really holding up.
Mac 00:30:48 Um, and actually, uh, in some sort of side work that our really talented postdoc DUI Mueller in my lab, uh, has done. Um, there’s actually a really interesting analogy with this, this type of diffusely projecting system, which is that they work a little bit more like temperature, uh, in the Woody and in a glass of water than they do like a message being passed and to cut a very long story short. Cause I don’t know if we want it to go there. Yeah. We can come back to it later. Um, the idea is that by having a kind of diffuse signal that passes up to many different areas, what you’re essentially doing is, is allowing those, those, um, contacted areas to become more likely to be part of some active coalition in the system. You’re sort of imbuing the system with a flexibility and a variability that it wouldn’t otherwise have so much the way that heating up a glass of water, lets a little water molecules whiz around in ways that they couldn’t, if they were stuck, let’s sander in a bunch of liquid or a block of ice,
Paul 00:31:37 Uh, is that Becca is the idea that I’m sending the diffuse diffuse projections is raising the excitability of the neurons or is it the actual firing of the neurons that puts them into a regime where they’re supposed to be more like ensembles, et cetera?
Mac 00:31:53 Yeah, I think it’s probably one of those little column, a little column B kind of things because the matrix elements, uh, not the only structure in the nervous system that have this sort of type of diffuse projection. And one of the other main culprits is an area that is probably kind of my favorite in neuroscience at the moment, which is the ascending arousal system, which instead of projecting up in that diffuse way and releasing glutamate like the matrix elements would, which we typically think of as that kind of message passing type, um, neuro-transmitter they released a whole other class of neurotransmitters that fit into this category called neuromodulators, which, um, they’re a little bit more like a whole modal style of passing where they, they, um, hit their little receptors, but the main impact they have is to change the kind of internal milieu of the cell.
Mac 00:32:36 They release some calcium or they open or, and voltage gated channels, and that can change the excitability of the system or the receptive, the receptivity of the system it’s in the way of putting it, um, that can kind of alter information processing modes can change the kind of, um, w we call this the kind of, you know, uh, sort of the state of the brain, uh, but kind of an inverted comments. Um, so yeah, I don’t know if we want to get to that. This is the problem with Tim, the neuroanatomy, right. It can go into many different directions.
Paul 00:33:03 That’s the thing, right. So yeah, let’s come back to the neuromod modulators because, uh, you’re doing theoretical work, uh, with that as well. And you know, you’re just interested in everything. So I, I would like, I know, but you know, you’ve got to coach soccer and stuff. How do you have the time, but no, so sorry I interrupted us there.
Mac 00:33:24 No, no, I’m just let me see if I can pick the thread back up. So, um, so if we were traditionally thinking about the cortex is special and everything else is kind of boring and we change our mindset and we say, okay, that everything’s been there the whole time. How has it elaborated? Now we need to start asking, well, how is the thalamus organized? And if we take this cortex first perspective, I think we can kind of lead ourselves into thinking the thalamus is just passing us a message. It’s sort of, you know, the Butler waiting at the front door don’t bother the cortex unless anything really important happens, right? That’s maybe one way we could start to think about it, but if we take Ted Jones’s perspective, we can think now the thalamus has this different set of capacities, a different set of ways of interacting with the nervous system.
Mac 00:34:07 Um, that might be really beneficial for explaining the kinds of modes that it can process in. And so if you’re interested in conveying that you heard a particular sound or not, maybe that core system is still really good, but if you want to make it so that you work out what that sound was, how could I disambiguate that sound from a whole bunch of different ideas about what that sound could be? I would argue that something like the matrix, system’s going to be much more helpful because what it’s going to do is it’s going to bring online way more systems that are going to help you to disambiguate and make the system function in a different kind of a mode. So that’s really the starting point. Um, but you know, once we’ve taken that first step away from this cortico centric perspective, it’s like this sort of endless walk, there’s not, you can’t help, but take that next step.
Mac 00:34:51 And if you look at the thalamus, um, one of the really, really interesting things is that even though it has these different blends of cells, it actually has different sub nuclei within it as well. It’s a really fascinating structure near anatomically. And we’ve learned a lot in the recent years from these really brilliant studies from, you know, groups like, uh, the Allen brain, uh, group and Adam Hammons group has this beautiful paper on the thalamus, um, looking at all of its complexity of the different projection types. Um, but one of the things that, um, you know, I’ve, I’ve become really interested in is thinking about, well, what other structures actually interact with the thalamus? Let’s call it from below. So if we, if the cortex is its connection to the top, how it’s interacting first order, higher order, or call matrix type projections, what kind of projections impact the thalamus?
Mac 00:35:38 And this is, uh, you know, again, one of these kind of interesting stories that shows you in a way, what the chefs doing in the kitchen, um, and it makes the chef look, uh, completely incompetent. But, um, I, when, when I hired my first postdoc, again, Eli that I mentioned before, um, because we came from such different worlds, he he’s from the world of physics and, you know, I have this background in medicine and neuroscience. We really wanted to kind of find a way to talk to one another. I, you know, I really wanted to hire someone that could help me do computational modeling at a high level. And Eli’s really got great experience in that space. Um, but in terms of his neuro anatomy, he’d really focused on the basal ganglia, um, which is a fascinating system, but I wanted to talk about a little bit more than that, this sort of bigger picture.
Mac 00:36:20 And so we had these really great conversations. And I think when you, when you have really great conversations with an another enthusiastic scientist, you can both find new patterns in places that you may not have expected them, uh, like that temperature analogy I mentioned, but you can also quickly realize you’re an ignorance. And I realized in trying to explain to Eli how I saw some of this fitting together, that I really didn’t know enough about what, how the, these different big structures in the subcortex like the basal ganglia and the cerebellum were impacting the felonies. Um, and the reason I bring both of those up is that, um, the thalamus, as I mentioned, has all these different structures in it. And in the, what we call the eventual tier, the kind of most interior part of the nucleus is a couple structures, the ventral anterior ventral, lateral nucleus, as well as the medial dorsal nucleus that received basal ganglia and cerebellar inputs and are known to be part of these loops, that kind of project back up to the cortex.
Mac 00:37:12 So, but we don’t really know a huge amount about exactly which regions projected, which, and there’s lots of different stories in the literature about this. And so this is something that, you know, I’d known really from back in my PhD days. I know that there was this uncertainty and, um, I’d had all these great conversations. Uh, Charlie Wilson at UTS was an absolute, absolute savior doing it during my, um, my post-doc, I was over in the U S working in California, but my wife’s from Texas. So we always sort of spending time both in San Antonio and in Palo Alto and at UTC, Charlie and I would have these fantastic long conversations about the basal ganglia and the thalamus. And that man is brilliant. Anyone who gets a chance to chat with them should buy him a beer. Um, anyway, we were talking about this and, and he put me on to some really great research from the group in Japan for promoter is the, is the first author.
Mac 00:38:02 Um, and essentially what they’ve done is they’ve taken, um, uh, rodents thalamus, and it looked at these ventral Tio and they tried to work out, which, uh, if we characterize these, um, different thalamic cells, according to some staining, which regions are projecting to the different types of stained filmic cells and to cut a long story short, it turns out that they stained the cells with exactly the same stain that Ted Jones had used to find the core matrix populations. And lo and behold, they found a really, really clear rule, which is that the core cells in the ventral tier, the ones that project precisely to the cortex, they receive glutamatergic inputs from the subcortex most predominantly from the deep cerebellar nuclei, the main output of the cerebellum. And in contrast, the matrix cells, the ones that project, a fuselage to the cortex received permanently GABAergic input from the Globus, pallidus the main output of the basal ganglia.
Mac 00:38:51 And it was like, it was like this sort of, you know, this moment where everything kind of went, oh, wow. So there’s this story about the thalamus and the cortex interacting that has this really clear kind of difference in terms of it has a population dynamics that could emerge from it that is receiving very different inputs from these different sub-cortical structures. So then all of a sudden it was like, when you’re doing a puzzle and you’re like, oh, that’s how it goes. You just like, turn it to the side. And it clicks in like, oh my God, that’s amazing. And I spent the better part of a year, you know, really reading and thinking about what the implications would be for that little twist, where we can all of a sudden have a discussion about thalamus cortex, basal ganglia cerebellum, and think a bit thinking about how those interact and what the implications would be for all these other mechanisms and stories we have that are really, you know, grounded, granted grounded, let’s say much more in the cortex or much more just in the cerebellum or much more just in the basal ganglia.
Mac 00:39:48 And all of a sudden we had this way of kind of talking about them together. And, um, you know, I see this as, you know, the first very gentle step in, in what I, what I view as a massive landscape of possibilities of trying to understand the system in detail with the kinds of tools we have now up to genetics and, uh, you know, really great, um, detailed characterization of these systems, as well as with computational modeling, I think is a really great opportunity to, to think about these things. But that was the paper we’re discussing is really kind of was the first step in that direction. I was saying, how might this stuff work together? And what might the implications be?
Paul 00:40:21 I love those moments in, uh, they’re very few and far between at least three day wear, but, uh, those moments where it, where it really clicked,
Mac 00:40:31 You know, that there’s a little side story here that I think is sort of fun as well. Um, so I, as I mentioned before, I did a PhD, uh, in FMI, but not the focus of my PhD was on Parkinson’s disease. And so everything in the world of Parkinson’s disease is basal ganglia and document the whole game. And we were trying to stretch that out a little bit and think about the system a little bit more because there’s a bunch of symptoms in Parkinson’s that don’t really make all that much sense in the traditional kind of opened the gate, close the gate in the basal ganglia story. Um, so it was already kind of thinking about things from that a bit more zoomed out perspective, but I feel like I’d, I’d read probably too many papers about the basal ganglia at this point in my career.
Mac 00:41:12 Um, but the cerebellum didn’t make as much sense to me. I didn’t really have as great of a feel for it. And, um, it was actually, uh, my father, uh, he’s, uh, an evolutionary biologist and, um, being an evolutionary biologist brains to him and kind of just like a boring side detail. That’s what interesting beauty of the ecology ecological universe. Um, and so having my father talk to me about neuroscience was quite, uh, uh, quite a trait. And one day, uh, when I was kind of towards the end of my PhD, he came to me and said, um, you know, your, your son Tyler at this time, Tyler was about year and three or four months. He’s like back about two or three months ago when Tyler was walking, I noticed that he was kind of stumbling around and it was really effortful. And he was really focusing on everything he was doing.
Mac 00:41:56 And now when I watch him run around the pockets, like nothing, nothing bothers him at all. He just runs around the world and what’s happened in his brain. And I was like, okay, number one, my dad asked me a question about the brain, like what’s going on right here. Um, but, but number two, it, it really kind of threw me for a loop because here’s this abrupt change that’s happened in my son’s ability to do things effortlessly. And I think that’s a really, really crucial feature of our, of our nervous systems, particularly humans, right. Where my dad likes to say that, um, humans are, we like to think we’re the kind of pinnacle of evolution. Um, but if you lined us up against any animal in the animal kingdom, they would beat us at the thing they’re good at. We’re not the fastest. We can’t swim that well, we can’t see very well. Um, but we’re, we’re really flexible, right? We’re born extremely organized and we can do things really, really, uh, we can do lots of different things wherever we put our mind to, we can learn, but we learned to do them quite effortlessly, such that we don’t really think about them anymore. And I think this, this case with my son learning how to walk, uh, with this kind of effortless nature was a really, really good example of this.
Paul 00:43:01 I thought you were going to say that you’ve told your dad, oh, dad, that’s the cerebellar dilemma, a critical loop coming online through the course. You didn’t, you didn’t just immediately say it. Yeah,
Mac 00:43:12 No, no, no, no. I did not. No, no, it’s, this is, I, I think this is important though, right? Because I think we have this conception sometimes in science that we, that, you know, think the clouds for kind of shine of light pops down and you’re sort of somehow inspired by something I wasn’t inspired by the answer. I was inspired by how hot that problem was. Right. How could we do this? And when I went and read the literature, most of the literature that talks about what we would say is sort of habitual behaviors focused on the basal ganglia. Um, right. We would say something like, as you learn how to do a particular habit, it’s something like the anterior pots of the basal ganglia decide that something is interesting. And then over time it gets passed back to more posterior parts of the basal ganglia and they kind of execute the habit.
Mac 00:43:56 But as I said, coming from this world of Parkinson’s disease where we’ve been thinking a lot about the basal ganglia, I was like, well, look, that makes sense that you’re you’re in some sense, but we also know that the basal ganglia is really a dimensionality collapser right. The number of cells in the striatum is a couple of orders of magnitude, less than the cortex. And again, you collapse it to the pallet and before going back to the thalamus. And so if you you’re going to store a really precise habit, it doesn’t seem like the best place to do it. Right. You you’ve, you’ve sort of, why would you waste your resource in this space of going, which is already limited on one very precise thing when there are other parts of the brain to do it. And so I was immediately a little bit skeptical of this story just from a kind of first principles perspective.
Mac 00:44:37 And then the more I looked, the more I read, I was like, oh my God, around the corner in this place that everyone likes to forget. And if you tell most neuroscientists about it, that aren’t studying it, they’ll yawn and pretend like they’ve got some other problem to solve, but come on, man, the cerebellum is just beautiful. I mean, it’s, it’s over half of the cells in an adult human body. Most of them have these little tiny granule cells. And it does the exact opposite of the basal ganglia when it receives an input. Usually the outputs of the, um, the layer five parameter on, so things sitting on the picture behind me from the cortex, it does a dimension out of the expansion. So it’s a little bit like the Colonel trick in machine learning, right? It’s like spread like a reservoir network. It’s spreading out the signal.
Mac 00:45:16 And now you can condition on all the little subtle differences that you might never have guessed were important. And whatever’s most adaptive gets fed back to the cortex as a little guests as what’s coming next. And that is the one that really hit me where it’s like, oh, of course, what Tyler’s learnt how to do is to anticipate what’s going to come next when he’s walking. So now he doesn’t have to think about it. He can just do it. And I think this translation from deliberate processing to more sort of what we would call delegated automatic processing is absolutely fundamental. And so a lot of the history of this, uh, the paper that we’ve been discussing is in trying to put some foundation beneath that insight, uh, and to try to work out what the neuroanatomical basis might be, um, of that kind of mechanism, just to kind of, to summarize the, I think if we think about just the kind of cortex we can come up with certain answers or mechanisms for how particular functions might arise.
Mac 00:46:14 Um, and so, you know, a really good example would be the predictive processing we talked about before, right. Um, if you have an expectation that something will happen, then the evidence that comes in, um, can interact with that expectation and give rise to what we call the posterior. And then you can kind of act on that. And this is a kind of really classical framing to think about, um, a lot of, um, psychological functions. There’s a ton of evidence for this. Um, but what we don’t know, and Michael Spratly’s work is really, uh, great and I’d point people towards, uh, towards his work to this, to this end, what we don’t know is how exactly that kind of a computation is implemented in the brain. And we have, uh, different, uh, hypothesis about it, probably the most popular, uh, it comes from an old idea from Robin ballad.
Mac 00:46:58 Um, and the idea there is that if you want it to be maximally efficient, the best possible thing you could do is to have a P uh, prior come down and evidence come in. And if they match don’t do anything at all, right. If you get a good match, that’s it, you’re done. The match is finished, but if you don’t get a match, then what you should do is you should send a signal to other areas to try to work out what the appropriate natural as they call that a prediction error, because you made a prediction. The evidence came in and there was an error. So that’s the way we typically think about that in a brain is that, um, the cortex sort of sends a little projection down. That’s the prior that the evidence comes in, let’s say via the thalamus and its little relay, the Butler passes the message to the, to the person sitting in their drinking room.
Mac 00:47:41 And then they kind of all look at matched and we move on. Um, the problem is when you look at the anatomy and you start to think about the rest of the system, you realize that while that might be one way that this system works, it’s missing some of the benefits you might convey by having the sub-cortical system play a role. For example, the thalamus can project up in that diffuse way and act as a, you know, what we would call a Pryor, but it doesn’t seem to have a lot of the other kind of constraints that the cortex has in its specificity. So if, if the thalamus projects up versus let’s say, you know, some area in the posterior parietal lobe, we might give those very different labels in terms of their functions, but they might have exactly the same kind of an impact on the system.
Mac 00:48:23 So that makes you sort of question a little bit, um, another is that the cerebellum is projecting up much in the way that the kind of evidence that would come in from the lateral geniculate nucleus of the thalamus is coming in. And so it just makes you question how these different systems might function when we kind of take that more zoomed out perspective. Um, and so I, I think, you know, again, uh, this is really the first step, um, in a, in a direction that I think requires a lot more research. Um, but trying to embrace that systems level perspective when we’re thinking about the functions that only emerge from the whole system, right? That our ability to have a conversation is not just because we have well-developed several quarter CS that can process language. It’s also in our ability to, uh, anticipate when the end of a sentence will happen or to know the right time to ask the right question to push us in a different direction. But it’s also in our ability to be awake and alert and to even process information in the first place. And I think as much as that might sound a little bit, like, you know, hand-waving, I think it’s important that we remember this and that we remember that the kinds of functions that we’re trying to describe and not due to parts of the system, but the system as a whole working together.
Paul 00:49:31 I think that was well put. Yeah. So, so that’s, so you have a role, um, I mean we can kind of step through a couple of the, uh, implications that you talk about. Did we, did we get enough of the story in of the basal ganglia versus cerebellum and how they, uh, put the brain in different modes and how we, one way to communicate that and think about it is through dynamical systems theory, or should we kind of summarize that before we move on?
Mac 00:49:59 Yeah, no, that’s yeah. Looking at, we can give that, give that a crack. Um, so yeah, over the course of probably the last, um, four or five years, uh, I’ve been really led down this, this path towards thinking about the brain as a, as a sort of constantly evolving dynamical system. Um, my collaboration with Michael break Spears out of the university of Newcastle has been really, um, uh, really inspirational in that front. And, um, and then working with Eli and Brandon, my two post-docs, uh, also Johan John and cardiac tryna, and Cal Sawyer. We have this little fun group that kind of integrates across some of the, um, this space to try to just have discussions about how can we, how can we frame some of these ideas that we’re thinking about in neuroanatomy in a dynamical systems, language and vice versa. Um, and so one of the challenges that I kind of set for myself with the paper we’ve been discussing is to say, okay, we go to this neuroanatomy, but if you’re not a neuroanatomist, I might be describing core cells and matrix cells and calbindin positive and pop Alvin and positive itselves.
Mac 00:50:58 And it might just be like, you know what, this isn’t for me. I don’t like to think about different cells. I want to think about what the implications are. And so I wanted this challenge of trying to think, all right, if we think about the brain as, uh, having these kind of, um, low dimensional structure to it, and we think about how that structure is evolving over time. I really great analogy for, for that process. And in fact, it’s a little bit more than analogy because the kinds of equations that would describe that analogy are really directly related to how these, the nervous system is interacting with itself. Um, so a really great analogy is this idea of the attractive landscape. And so the concept would be something like this. Um, if you conceptualize all of the billions of neurons in your nervous system as a little boy, a little bowl, um, the way that that nervous system will interact over time, or the way that it’ll change over time is dependent on what kind of opportunities are present, present to it.
Mac 00:51:50 Um, so you can imagine, you know, um, you’re sitting in front of the coffee machine in the morning and there’s a little button on the coffee machine in front of you, and you have to push it to make the thing up and say, you can put the coffee party in, or if you’re really fancy, you have to like put the coffee in a tamper down and do all your history stuff. I have no time for that. That’s the secret to my time to read is that I use an espresso. Um, so, um, you have this opportunity to present to you. And the way you might think about that is that there’s an attractor or a basin that’s pulling the system towards it. Um, and so this is a traditional way of thinking about kind of dynamical systems frame. It doesn’t have to be a brain, it could be any kind of dynamical system you’re interested in.
Mac 00:52:26 Um, but for, for me, I was trying to think about in this way. And so I started to think, well, if you’ve got an attractive landscape and there’s a really, really deep, well, that’s a little bit like a little coalition of neurons firing a lot, they’re kind of basically saying this is where we’re going to move towards. And if you’ve got a lot of neurons firing at the same time, that’s a little bit like saying, oh, there’s a thousand different options in front of me. I don’t know which one I should go towards. And that’s a little bit like the landscaping Flato and we have some computational modeling that kind of fleshes out, pushes this analogy out in a much more satisfying way if you’re interested. Um, but then I started to think, well, if the basal ganglia is projecting to these matrix, uh, population, right, they’re really diffusely projecting one.
Mac 00:53:05 And the cerebellum is projecting to the coal really precisely projecting one. Um, then maybe they’ll have different, uh, impacts on how the brain state, at least in the sort of ventral tear of the thalamus to the sort of frontal cortex might evolve over time. And so if I’m coming across a particular, um, challenge, like, you know, uh, how do I, you know, shoot a basketball? That was a lot of my intuition was this. We just had a basketball hoop installed in my backyard. I was trying to remember how to shoot free times. And, uh, it didn’t start out particularly elegantly, but I’m thinking like, what am I doing here? Um, but what I noticed with that is that early on, I tried lots of different things. Maybe I’ll talk to my elbow in, maybe I’ll kind of flip my wrist more, maybe I’ll turn my shoulders to the side.
Mac 00:53:43 I had all this variability in my system theater important. You gotta get the feet pretty straight, sorry, the feet that’s right. Yeah. You got to find it. But then in the NBA, like everyone’s always jumping around with their fate and stuff and Steph card sort of flicks the ball. Like he doesn’t care about it. It’s crazy. Right. Um, anyway, I was trying to think about this and I was thinking, well, if the matrix thalamus is receiving the basal ganglia input, and I know it’s involved early in learning, uh, and it’s gating via these sort of two inhibitory populations, the, the eight inhibits the stragglers powders, which then inhibits the Thelma. So what you can do is you can turn off that inhibition. Well, that’s gonna do is it’s gonna release these matrix population, uh, just a little bit. And what that’s going to do is it’s going to come up and it’s going to kind of diffusely sprinkle activity across the nervous system.
Mac 00:54:27 So it’s not just going to release one little plan, which is kind of the traditional way I’ve been taught about the sort of these segregated basic, um, ganglia thalamocortical loops. But rather what it’s going to do is it’s going to basically flattened the landscape, but just in that one little particular location. So if I’m shooting a free throw and I miss 10 times in a row, maybe it’ll just let me tuck my elbow in. Or as Paul said, move my feet to the side in contrast as the cerebellum balloons, every action that I do every time I make a movement, there’s a little reference copy. That’s coming down to the ponds, shoots into the cerebellum, does its expansion comes up to the cerebral cortex back to the date on the nuclear, back to the thalamus, to the cortex, right? That loop is going to be saying every time you make that move over time, I want you to learn that particular one that led to you getting the shots in and ignore the ones that you did.
Mac 00:55:14 Now, the caveat here is that the cerebellum looks like it loons in a slightly different way than the, of anger. It’s not learning via reward. Prediction. Error is, um, it’s, it’s a supervised learning where th th there’s a, a template and a, and an input, and you try to match them together. But with that detail aside, the cerebellum Laurence has sort of take over that function and do it in a really precise way. And so in the language of dynamical systems, we could think, well, if there’s an attractor and the basal ganglia gets involved, that’s going to sort of flatten it out a little bit. So that if, if option a was the one you were doing, now, you could try B or you could try C. And if you’re using the cerebellum, if Amy shows up and that’s the start of your sequence, let’s just get to pee real quick and then see an end date.
Mac 00:55:52 And then it’s going to let you quickly move through the sequence and a really automatized way. You don’t have to think about it anymore. So if I get all of that movement down now of a sudden, I can have a conversation with someone while I’m doing shooting free throws, or I can be thinking of a project in some other part of my brain while I’m actually executing really autonomously. And that’s where this idea, I think, um, you know, links back to some really fascinating psychological phenomenon that we don’t have great explanations for in the brain, like how you can be driving home from work and completely forget that you drove home. You don’t even notice it the whole time. Right? You did that job expertly in that context. Um, but you, some, you somehow weren’t aware of it. And so there’s all these really, really fascinating, um, sort of mechanisms out there in the brain like the cerebella loop, but I think are really important for some of those functions, um, that I think, you know, maybe this moves us a little bit closer to me, how to make hypotheses about how we do things as are the automatically, um, like the driving while talking.
Paul 00:56:47 Yeah. I mean, so you talk in the paper and speculate about system one and system two and give roles for the cerebellar loop and the basal ganglia loop there as well. Then maybe, maybe you could just brief summarize that and then, uh, I want to bring it home and ask you a few questions about the thalamus in particular.
Mac 00:57:04 Oh yeah, sure. Um, so the system wants system to kind of label is sort of part of a broader group of what we would call Juul process theories. And these are really, you know, have been around for a really long time. Um, Shifrin and Snyder did a lot of the really great early cognitive neuroscience work. And they’ve got these two papers that were like both a thousand pages long and full of experiments. Um, but, uh, Gordon Logan also did some really cool stuff with the sort of inner loop out loop, uh, ideas. So this is a sort of an old idea, but Conaman in his sort of brilliant way and kind of really stepped down this really kind of lovely kind of label for it. Um, and the idea is that, um, there are kind of different modes in which we can interact with cognitive problems.
Mac 00:57:42 Um, a lot of the time, in fact, the vast majority of the time, where in what he calls system one, which is where we have preconceived notions for a particular, let’s say context, and let me just act really quickly. We use what he would call a kind of heuristic. We say, if the context maps something that I think I’ve kind of already experienced before, I kind of know about, I’m just going to act in this particular way. Um, and a lot of the time, I think we take this for granted, it’s actually really useful in a lot of senses. Think about how many systems, one type functions you have for language, for example, in your language that you understand, you can pause almost any kind of word, any kind of phrase. You can quickly work out if someone’s used a verb around the wrong way with a noun.
Mac 00:58:22 Whereas if you walk into a foreign country where you’ve never heard the language before, it absolutely sounds like, uh, you know, experiments will be off musical. Something like you just can’t pause anything. Even the, even the cadence is really foreign. Um, so I think systems one is everywhere and it’s sort of like our default way of interacting with problems. And if that’s not working, if systems one can’t handle things well, um, or we do it in a deliberate fashion and agency is a whole other question that you probably need to talk to way smarter people about. Um, but if we do things in a deliberate way, that’s like kicking the system up into what we call system to like, which is really deliberate, much slower and much more kind of conscious processing where you’ve got a focus and you’re really kind of doing work on that.
Mac 00:59:06 Um, and when I was thinking through some of the implications for this sort of neuroanatomical perspective, um, that, uh, that, uh, that I’ve been speaking about, one of the things that I think is a sort of intuitive way for people with that background in psychology to kind of make contact with this is to think of, to a first approximation, these kinds of cortico cerebellar cortical loops. There’s a little bit like systems, one, like you have a particular action and you sort of jumped to what, what to come next. Whereas the basal ganglia system is a little bit more systems too. Like it’s sort of allowing you to focus on one particular part of the system, um, and, and really sort of drill down on it. Um, now where it gets really interesting is that, you know, I mentioned before that the basal ganglia has this sort of, um, contact with the matrix elements, right, which produces this diffuse projection to the cortex.
Mac 00:59:50 Um, but what it does up there is not necessarily just sort of sprinkle activity everywhere. What it’s actually doing is it’s activating the apical dendrites of the, uh, these really massively a five parameter neurons. And this is where the story really hit home, hit home for me. Um, there’s some really beautiful work. Matthew lock-in, uh, is sort of one of the pioneers of this, but there’s many others that have shown that in the cerebral cortex, the apical dendrites of these critical parameter ons are actually separated away from the cell body in a really fascinating way. It’s almost like they’ve been pulled apart. Like you’re sort of stretching a band to make them as far apart as possible. And what that means is that you can actually kind of do contextual processing on these. What you can have is the context that comes in to the apical dendrites changes the firing properties of the cell body, but only when it exceeds some threshold.
Mac 01:00:38 So you need to have lots and lots and lots of inputs, or the neuromodulatory system has to come in and close a little leak channel that allows the system to now fire in a different kind of a firing mode. And so what that means is that the basal ganglia can increase the variability of the system, but still create the action of the system. It can still allow the winning coalition in quote unquote to fire and have an influence over the next brain state and to sort of enact that change. And I actually think this sort of actually comes back to the predictive processing, sorry, a little bit, one of the parts of the round valid model that I find, um, a little bit counterintuitive is that if you get a good match, you just sort of don’t do anything. So if I expect your voice Paul, and then you speak and I get a match, I should just sort of, you know, get out of the way everything’s fine.
Mac 01:01:25 But to me to be adaptive, that system needs to act on the thing that they guessed was there and actually was right. If I guess that a saber tooth tiger runs through the, through the door and it actually does. And I just sit here going, oh, good. I made a good prediction. You know, there we go. Don’t do anything I’m cooked. Right? So what I’d like to do is have a system that can enact that change and get moving. And the beauty of Matthew Hawkins, a cellular model is it allows that prior that hits the apical dendrites and the evidence that comes in to the, the cell body to actually make a difference to cause the cell to perse fire, to then be the one that goes down to the cerebellum and says, what, what, what should I do about this to go down to the pencil canyon and say, what are my options right now?
Mac 01:02:05 So it kind of bakes it in together in this way that you can see that the elements of the, of the round valid model are undoubtedly related to the nervous system. You do look like you make priors and look at evidence and match them up. But maybe what we do with that information might a little bit different than we originally intended. And we can think a little bit about what that might mean for some of our models and, and, uh, you know, again, this is a step in the direction where we need to do a lot more empirical work to really constrain these ideas. These are really theoretical concepts, but it’s exciting time because I think we have the tools to start to do these experiments.
Paul 01:02:38 So I’m going to ask you a question that, uh, I feel is in danger of, um, I’ll be, maybe you can laugh in my face because we’ve just been talking about how important the whole system is, right. And to consider the interactions of the system, because it’s really the emergent properties from that. That’s important. But what I’m going to ask is the bottom line of how to think about the thalamus then. So should we think about the basal ganglia as always going full, full bore and the cerebellum has always going full bore and the thalamus controlling them, or mediating them, or nudging them to change that dynamical landscape to put us in different regimes of action and thought, how do you think about that thalamus in that role?
Mac 01:03:20 Yeah, that’s a really great question, Paul. Um, I, I think part of the problem is that I don’t think the thalamus is really just one thing at the end of the day. I think it’s in some sense, it is, it, it ha it has a, you know, a particular typology where there’s a bunch of glutamatergic cells that don’t contact with one a lot, and then a bunch of inhibitory cells that compete with each other from the RTN and, and it protects to the cortex and receives input. So in some sense, you can think about it under the single label, but in another sense, when we drill down into the details that that kind of unity starts to break apart. And so I think it really depends on the question you’re asking. Um, if we’re thinking about the coal thalamic nuclear, I do think that the message passing story’s a good enough first approximation to think about it.
Mac 01:04:06 Um, with the matrix alignment nuclei, we like to think about it a little bit, like changing the brain state, um, kind of, uh, sort of increasing the excitability of a population, but that’s, again, a first approximation. There’s a lot of subtlety there. Um, one connection we haven’t talked about, which is, again, is systematically under studied in the literature, is this massive projection from the thalamus, these matrix plot population, very particular ones called the intro lemonade nuclei, like the parafollicular nucleus and the central median. They actually project really strongly to the striatum, uh, and, and have a really strong gaining influence over both the spiny projection neurons in the culinary getting ruins. And that structure, that interaction is something that we really don’t have great empirical work on. There are people working on it. Um, but it’s, it’s an area that needs to evolve for us to really understand what’s going on.
Mac 01:04:53 It’s really important structure. So overall, I don’t think that there’s going to be sort of any one label, but, you know, th th the, the, the kind of picture you were painting just before in your summary, I think is a good way to think about it. Right? One thing we do know about the thalamus, um, is that it it’s, um, has stock changes between sleep and wake or between Anastasia and wake. So Messiah, um, really convincingly showed, you know, 20, 30 years ago, that if you look at the foreign population of filmic neurons during sleep, there, there’s a particular calcium channel that’s closed. And then when, when you wake up, it opens up, well, maybe I got that role. It could be open and close up, it’s a switch. And then when that changes, you can then get a confirmational change in the kinds of interactions that can occur.
Mac 01:05:38 And all of a sudden you get this emergence of high-frequency desynchronized cortical, EEG. Um, and so I think in a way what the thalamus is doing is really controlling the state of the system, right? And then any influence that happens to it all the time, whether it be a cortico influence, a basal ganglia influence, cerebella a curricular influence, a neuromodulatory influence is going to shape and change the way that the state will change over time, which is one of the most crucial factors for determining how we do what we do. So I think of it as an absolutely core pot of, of the central nervous system and, you know, really, you know, at a kind of big picture level, what I was trying to do with that paper, a lot of what I’m trying to do in my research program is to sort of shine light on these kinds of, uh, you know, areas of, of the nervous system that we haven’t really thought about from that kind of systems level perspective, as much to sort of show that they’re playing a really crucial role. Um, but you know, it’s going to be up to people like Michael lasso that have all the, like really crazy positive genetic tools that let us really go in and, and, you know, not out the details of these circuits that I think will actually really ended up carrying the day.
Paul 01:06:45 Okay. That was satisfying. Thanks for not laughing in my face. I mean, I, I think that I still have this, um, bias to think about a controller to think it was somewhat like of a homonculus right. That, uh, there’s these different states in the brain that they’re being switched. How are they being switched? There has to be a controller. Oh, maybe that’s the thalamus, but on the other hand, if you consider like the whole systems level processing, uh, it’s less satisfying to say self-organizing, but I don’t know. Do you think that, that we need to think about it in that respect as a self-organizing system that we’ll just have to accept it? Uh, not that that’s disappointing, it’s just, uh, I think people like me with a simple mind, tend to think in a homonculus fashion and think something’s in charge, you know?
Mac 01:07:31 Yeah. Um, you know, uh, I, I don’t remember who said this quote, but if, if not, we should just give it to ourselves Paul, but biology is weirder than you could ever imagine. Right. Um, so I think this comes back to the fact that, um, I think once we embrace the, the use of neuroscience and we, we, we really kind of put to the side, any concern we have about trying to deeply understand the thing just yet, let’s just get heuristic approximations, and then we can go from there. Um, I think once you, once you take that perspective, you start to realize that making some, uh, discovery or having some hypothesis makes you make commitments to the shape of the system or the way it interacts in ways that you may not have anticipated. And, uh, the dynamical systems stance is one that I think was kind of inevitable when I started reading more about neuroanatomy, because you need that language to understand how these things are interacting with one another, because lack of the brain is a complex system that has lots of interacting parts.
Mac 01:08:38 And the main thing they can kind of do is nudge one another, or kind of increase the excitability or the receptivity of each other over time. They, they can’t do anything about the past. Well they can do is about the near future. Um, and so in a way you need the dynamical systems language to kind of help you, uh, even frame the things that you’re seeing in, in, in sensible ways. And one of the things that I think comes with the dynamical systems language is this self organizing concepts and really weird things like there’s a term called circular causality. And I was talking about circular causality back in the day as like a slap on the wrist. Like, don’t do this because you’re making the wrong, you’re putting the cart before the horse. And you’re assuming your answer right? What it means in this language.
Mac 01:09:19 Alicia, Gerard has a really beautiful book about this called dynamics in action. And what she argues is that there’s different kinds of causality happening and the bottom, you know, the, the kind of traditional billion bull style causality that we really like to think about where one neuron contacts, another neuron, and then kicks a message onto another neuron and onto another neuron. And there’s like this little line of chain of command, kind of a thing, maybe true in some sense, but there’s also this whole other set of what we would call causal rules, where the constraint of the system, the top level configuration system changes what’s possible for the lower level. And this is a really kind of gnarly concept, but one, um, you know, really intuitive example that I think kind of helps to kind of, um, play the sounds from George Ellis is that the kind of software that you’re running on your computer, whether it’s, you know, Microsoft Excel or, you know, um, Microsoft word, the same key stroke can have a completely different effect on the electrons running around in the hardware of your system.
Mac 01:10:16 So you can have a constraint from the top, the program that’s running really changed the kind of way that the system can then evolve over time and the kind of way that the different billable logic plays out. Um, and so, you know, I, I didn’t pretend to be an expert on this stuff, but I’m reminded of, you know, like the episode you did with mock on inter into activism. You’re thinking more about a systems far from equilibrium, trying to figure out how to kind of navigate a complex world, take the affordances available to them, to, to solve ongoing problems that pull them a little bit further away from equilibrium than they’d like to be. And that kind of a system is going to have these really weird features of self-organization. And we had dynamical evolution over time, and this circular causality that we need to understand better if we want to describe them in the way that they are, rather than the way we would like them to be without kind of traditional kind of a, to B to C causal models.
Paul 01:11:09 So thinking about your medical background and, and causality and causes and complexity, do you, what do you see as the prospects for being able to nudge these systems, um, in a therapeutic way and, or, you know, in a clinical way, right? Like, uh, are we close to being able to, like, let’s say your theory is completely correct. Right. Would you feel comfortable getting in there and pushing things around, uh, to treat, uh, people’s personalities?
Mac 01:11:42 Um, yeah, this is a great question. And one that, um, I feel guilty about a lot, particularly when my mother reminds me that I, uh, left a job in medicine to work in academia. Um, my father was quite happy about it. Yeah. He, my father said, well, look, now you’re a real doctor when I got my PhD.
Paul 01:11:58 Oh, wow. That’s the opposite of what my grandmother says.
Mac 01:12:03 Um, my wife always teases me as well that, um, if I, if I get sick of academia, I can go back and work in medicine where I can actually make a real salary. Um, but
Paul 01:12:13 Yeah, podcasting,
Mac 01:12:15 Yeah, it’s, it’s a really hard problem. And I think it’s, it’s worth again, you mentioned that this is a, the first step in what I imagine is quite a long path towards finding out the nitty gritty details. Um, but yeah, I think I’m excited for the opportunities that will be available in the coming years for areas of medicine that have traditionally not had really kind of, um, uh, treatments with punch, right. We, one of our best solutions, if someone shows up into the clinic with a thought disorder is to just sort of block all the catecholomines in the brain. And, you know, they don’t really have a great life after that. They eat too much because they can’t control their appetite and they don’t really have any ability to be, um, quite creative or interact with people. But, you know, at least they’re not having hallucinations.
Mac 01:13:05 Um, and so I look forward to trying to work out how we can come up with better solutions for these kinds of folks. And, and I don’t think, uh, at the end of the day, it’ll be just about, you know, stimulating the area at the right time. I think that a lot of what makes our nervous systems so fascinating. And again, it’s one of these problems it’s in the really too hard basket is that we really are age antic, uh, organisms. We, we do things we act in the world. And part of the challenge, I think is that with the nervous system, it’s like that on steroids. So if you give someone a treatment while that might set you off on the right track, you could just as easily be the thing that your nervous system works desperately to avoid. And it could be the, the, um, you know, the downregulation of particular types of receptors or something like that that ends up carrying the day and having confirmed clinical benefit rather than the primary treatment option.
Mac 01:13:57 Um, you know, deep brain stimulation in Parkinson’s is another great example, stick electrode in this part of your brain, turn it on with really high frequency. And all of a sudden things get better unless they don’t. And if they don’t tweak it a bit, and we really fully understand exactly why it gets better. And so I think any sophistication in our appreciation of the kind of workings of the nervous system at that systems level is going to lend itself towards better suggestions for therapy, but they’ll ultimately be about dynamic into interventions rather than just like one-off gave you the tablet. Don’t worry. You’re good. Now, you know, in a way the immune system’s kind of coasting, right? You have a vaccination, uh, and then all of a sudden your immune system can kind of cope with the next insult. I think the nervous system has a lot of the elements of the immune system in it. It’s very variable. It’s really good at dealing with, uh, finding patterns in the environment, but it’s much more dynamic in a, in a, in a specific sense where you can’t just kind of like drop a rock in the pond and then hope that that solves the problem. And I think it’s going to be much more dynamic in the future.
Paul 01:15:02 So Mack, you, um, spent a lot of time learning anatomy and the connections and, um, thinking about, uh, structures besides cortex, for whatever reason. Uh, and then you weren’t satisfied enough. So thinking about the whole system, now you’re bringing in neuromodulators and neuro-transmitters into how they interact at the systems level. First of all, what the hell is wrong with you, man? And secondly, uh, what’s, what’s the story there what’s going on with, with, uh, your current work on, on the neuromodulator neuromodulatory system?
Mac 01:15:34 Yeah, so th there’s a, a kind of another fun story, uh, to kind of unpack the, um, so during my post-doc, I was working with Ross Poltrack, uh, at Stanford on, you know, functional MRI, um, and trying to think about it from the systems perspective. So when I started working with Ross, the kind of really hot topic in the field was what we called dynamic functional connectivity. And there’s a lot to unpack in that. And none of the terms really quite, uh, capture what they’re supposed to catch up, but the concept was something like, instead of looking at the correlation between two blood flow time series over, let’s say a 10 minute window, let’s unpack it into smaller windows and then see what happens over those smaller windows. Let’s see if there’s any fluctuations such that there’s not a sort of stationarity to the system, but rather this kind of interesting dynamics and the tack that we were taking was to borrow a really beautiful idea from Kymera and Emeril, who did this really lovely analysis of networks.
Mac 01:16:31 There were metabolic networks and that of 2005 paper, um, where they basically took a metabolic network co-expression of particular, um, metabolic byproducts. And then they looked at the interaction between all of those as a network. And then they said, well, this is a really hard thing to describe. It’s really not only depends on how you look at it. It’s really multi-dimensional. Maybe if we summarize it into a bunch of little communities, we’ll run some kind of a clustering on it to find tight little, little communities or what we call modules. And then when we have that information, let’s start to ask, well, relative to that modular breakdown, how each of the different metabolomics kind of related to the whole system. And it turns out that one of the, the framing that they use at least is called a framing. So you calculate something like the, between connection, how much was an individual metabolic, uh, signature, like the rest of the system versus its own little group and a local one, which we call the module degrees, each score.
Mac 01:17:21 That’s basically telling you how much, like you were your little group, you were a module. Um, and what we were trying to do basically, was to take their framing and to put it onto these little dynamic networks, quote, unquote, that we were measuring, right? So we did, someone’s lying in the scanner. You get a bunch of data, you break it up into little chunks, you calculate one of these networks and you look at the configuration over time. And one of the, you know, through a long, long process, that involved really great questions. I mean, one of the things that I love about Ross and his group is that there’s a real kind of poignancy of really getting to the bottom of problems and trying to kind of not fool yourself, uh, which I think is really, really easy to do in a, in a big space like this.
Mac 01:17:59 Um, and in fact, it was two of his grad students in a lab presentation that pointed out that we had brought along this, uh, old, um, way that they chopped up the data in the original experiment. They basically like set boundaries and then CA um, characterized parts of this little space into different bins. And we were just tracking them over time. And the grad students both said, why don’t you just get rid of those bins, just look at the thing over time. And so we made this little histogram at joint histogram and then watched a movie of it. And it was just like, you know, a ton of bricks. There’s this big, massive fluctuation over time, uh, between these extremes of a really interconnected system with lots and lots of those with between connections. And then it really isolated system with these little, these little within connections, and we call those an integrated and a segregated network.
Mac 01:18:43 Um, uh, and so that was amazing, right? Wow. We found this really interesting thing. And then like, well, what does it mean that we were stuck with this really hard problem, right? It was like, all right, back to the literature, you know, it was like a common theme here, right? You, you find a thing and then you go, what could it be? And then you spend a bunch of time, you know, it’s important for the students to realize this, you spend a bunch of time meandering around the world, reading a cool paper, being inspired by some weird question that your friend asks you or your dad asks you about your son standing up effortlessly. These things often strike you when you least expect them. But I think there an underappreciated aspect of, of science, or at least the part of science that I really love of that kind of wallowing in your uncertainty until it resolves itself, I think is one of my favorite pots.
Mac 01:19:29 Um, and so we spent a long time, I remember a bunch of the postdocs in the lab being like, dude, when are you going to like, work out what this is? This is kind of getting boring with, with tight of you talking about integration, segregation everywhere. You see it. Um, and, uh, so Russ and I decided let’s go meet a bunch of different professors around Stanford. This is again, one of the benefits of being in a place like Stanford is you just walk down the road and you’ve like, you’re meeting up with like a brilliant economist or a brilliant information theoretician. Um, and, uh, it was actually an economist that, that, that put us on the right track. We were talking to Matthew Jackson, he’s done a lot of really interesting working in networks, uh, in, in, uh, in economics. And I asked him, you know, Matthew, are there any pots, any ways in, in economics that you can cause the system to kind of do a configuration or change like we’re seeing, could you like, you know, uh, remove tax.
Mac 01:20:17 And then all of a sudden everyone goes out and spends money and they all look like one another, or could you like, you know, boost people over here with a little bit of money and if you boost it just the right people trickle down, economics might happen in the system. Might, you know, if that was a real thing, um, the system might change. Um, and, uh, and then he said, no. And I was like, oh man, I really wish I had a great answer for this. But then he said, surely there are parts of the brain that, you know, they don’t have to be big. I can just sort of project kind of to the rest of the brain. And then they could kind of change how the different parts interact. And honestly, it felt like the old Monte Python sketch where someone slacks on place at the face of the fish, I was like, oh my God, of course, it’s the ascending arousal system.
Mac 01:20:56 Right? Of course that’s, what’s doing this. And so we then went and did some empirical tests and did some computational modeling with Michael break spear. And I’ve now, you know, I’m working on doing energy landscape analysis with my extremely brilliant, um, postdocs from physics and desperately trying to understand what they’re doing, uh, to try to work out what is the best way to kind of analyze the system, but at its core, I think the ascending arousal system is just so important for shaping the dynamics of the rest of the system. And so, um, you know, we probably don’t have time to kind of get into this properly, but, um, the kind of cliff notes are that instead of the, um, the system using glutamate and GABA as its main neurotransmitter, which we kind of think of as, you know, either starting off an action potential or quashing one, um, the main, uh, effectors in the system are more like a hormonal like structures that, um, uh, actually, uh, often derived from amino acids. So actually this is this the point for my, um, my dad joke, Paul, um, what, uh, how do you know that, um, neuromodulators, uh, really rude
Mac 01:22:09 Cause their amino acid.
Paul 01:22:13 Oh man, thank God you did the dad joke.
Mac 01:22:19 Um, I’m a dad I’m allowed to make drinks. So yeah. So you’ve got these, um, these really highly conserved systems and your brainstem and full brain that taken amino acids from your diet, or they use the byproducts of the Krebs cycle, uh, and they convert them into little intermediaries that then go off and have little, they have receptors that are act like a kind of little lock and key mechanism that they’re called G protein coupled receptors. And the G protein, couple of steps are really different than the AMPA and the MDA receptors that we typically think of rather than letting ions sort of shuttle into the cell in particular ways, what they do is they create confirmational change in the internal state so they can release calcium, they can open voltage, gated, ion channels. They can also do all kinds of cool stuff. They can act like transcription factors.
Mac 01:23:04 In fact, there were a couple of recent papers where they showed that molecules like serotonin and dopamine actually can bind to the DNA like epigenetic modification and can change the likelihood of an animal ultimately recovering or not from an addiction. Do they treat, I mean, it’s, it’s mind blowing stuff. Um, and when, so basically this system has a really different effect on the nervous system than the traditional kind of glutamatergic, GABAergic, um, effects. And so one of the things that we do a lot in my lab is try to think about what the different subtle differences amongst those different neuromodulatory systems, how they interact, what it would mean to release acetylcholine here, but noradrenaline there, you know, what is serotonin doing? And does it oppose dopamine or does it actually work with a, we, we try to ask these kinds of questions about the neuromodulatory system in its details and how it interacts with the rest of the nervous system. Because again, it’s going to have those implications for the state changes over time, the flattening or the deepening of the attractor landscape, or the ability for this coalition of neurons to form and then be alive and around long enough to interact with another system that you need in order to solve the complex problem in front of you. So I think of it as quite quite important for those systems level things.
Paul 01:24:12 So again, uh, so I’m assuming the thalamus controls all this right now. I’m just kidding. But again, we have to think in terms of self-organization and complexity or, or, you know, at this point, actually, you know, thinking about the arousal system, we also have to think about life processes and metabolic processes, right. Which is a part of the whole integrated system.
Mac 01:24:35 It’s part of the fun, right? Yeah. So these, so these, these systems are, um, you know, feedback and a massive feedback control from both themselves as well as a number of other structures. The hypothalamus is the, the key, the key one, but other areas like the Habana ULA really important, the para Dr. Gray. Um, and then, you know, there’s also, you know, cortical projections to these systems, um, that, or, you know, uh, basal ganglia in the case of dopamine and serotonin. So these systems that you can think about them as sort of having this massive confirmational change in different arousal states. So for example, um, I talked before about how the thalamus wakes up, uh, uh, when you wake up, um, a lot of that is due to acetylcholine that comes up from the lateral dorsal tegmental in the, in the brainstem that kind of kicks off that confirmational change nor gentlemen plays a role as well.
Mac 01:25:28 Um, and so you could think about these things as, as having this massive change that changes the arousal state, but then subtle fluctuations in those neuromodulators can also change what you can do right now. So one of my favorite examples from the literature, it comes from Susan, Sarah, and Sebastian Baret, and they talk about what they call the network reset phenomenon with noradrenaline or norepinephrine for your north American listeners. And the idea here is that, um, if you, if you imagine a widespread system that can change the receptivity of the system, what we call the gain of the system that can kind of make it more likely for a spike that comes in to get propagated, to, to have some meaningful output. Um, one way that that can be really beneficial for an organism is if you’re, if you’re sitting in the, in the scrub, you know, looking for bugs, your little marsupial, and then you hear a rustle in the bushes, if you’re too zoned in and focused on your meal, you won’t notice that there are predators snuck up on you.
Mac 01:26:20 It’s going to come and jump in at you. Whereas if you have that big burst of noradrenaline, now, all of a sudden, whatever it is, actually in your environment, rather than what you want to be in your environment, the food and exploiting that food. Now, all of a sudden the system is now susceptible to whatever the most salient signal is or the most, um, uh, most, um, important for your adaptive, ongoing life, um, can then carry the day and you can react to that news, um, stimuli rather than the one that you were reacting to in the moment. And I think this kind of flexibility is, is kind of informative for the kind of more adaptive stories that we need to be thinking about, to understand how a nervous system could benefit an organism over massive swaths of evolutionary time. Like if you don’t have that system in there, if you don’t have a system in there that can help you focus down on something like we think maybe acetylcholine helps with or something that might help you work out what’s valuable or not in a really complex kind of, uh, sort of space or a temporarily extended landscape that we live in, we think document might help with, and you’re not going to act in the most adaptive way you possibly can.
Mac 01:27:20 And so I, as someone who’s really interested in that phylogenetic perspective, I think neuromodulators play an extremely crucial role in that process. Um, and also, you know, um, a massive cytopathology in both the developing brain, as well as in your generation, uh, and a bunch of psychiatric conditions. So I think there’s, there’s a lot more to be learned about this space, uh, that I think would really help us to be thinking about that sort of systems level interaction as it plays out.
Paul 01:27:48 So Mack, I don’t normally have such a, a systems neuroscience, heavy conversation on the podcast, and you’ve taken us on quite a tour, um, over different brain regions and how they interact and the function and the anatomy. And now the neuromodulators, uh, the, the portion of my podcast audience who is in the AI research world and or industry, uh, is probably feeling helpless and lost, let alone many neuroscientists. Right. So what I want to ask you is what do you see when you look at a deep learning network or, and, or like a reinforcement learning system or a deep reinforcement learning system, how do you think about the modern, uh, AI approach?
Mac 01:28:32 Yeah. Um, I apologize if, uh, if the content of the podcast has been to,
Paul 01:28:38 Um, people people need to get, yeah. People need to get slapped upside the head with some facts and some theory. Yeah.
Mac 01:28:46 Yeah. There, there are pretty pictures in the papers if you, if you’d like to see some of them laid out a little bit. Um, so yeah, look, um, you know, I think AI is a, is a fascinating field. Um, and you know, I think I sort of sympathize with some of the people you’ve had on podcasts. So think of it as almost slightly orthogonal to kind of neuroscience in a way, and that it it’s sort of grown into its own fascinating space with lots of idiosyncrasies that, uh, not particularly inspired by the brain, but don’t need to be, they just sort of are the fact that it works in the way that it does. So in, you know, in my, in my day-to-day. So I think of them as sort of orthogonal. Um, I do see them, they’re being really great opportunities for communication between those, uh, those two spaces, you know, um, Blake Richard’s work is something I absolutely love where he’s thinking about those layer five permanent, or as I was talking about before and credit assignment, and trying to think about how you could have, you know, messages pass up the hierarchy, but then also learn which ones I should reinforce and increase the connection strengths between which ones I shouldn’t.
Mac 01:29:48 And I absolutely love all that work, and it’s not something that I, uh, I have a lot of experience in. I love reading that literature and thinking about how to kind of, um, integrate the ideas.
Paul 01:30:00 But what about the idea of, so I’m thinking about deep reinforcement learning in particular and heavy on the reinforcement learning idea here, you know, you have people like David silver, uh, and his colleagues writing a paper. I think it was called reward is enough, but essentially the claim is that all we need is reward in a reinforcement learning or deep reinforcement learning system that is going to lead us to AGI, right? So it’s not just tools that AI is after there. Uh, there’s a certain sector, uh, that is optimistic and interested in building quote unquote, true intelligence, whatever the hell that is. But, um, do you see that as orthogonal or do you see like a reward system as in enough?
Mac 01:30:50 Um, yeah, it’s a great question. I’ll have to have to read the paper. I haven’t haven’t come across it. Um, I I’d come back to something that I think Blake Richards actually said on your podcast way back, which, which is this lovely analogy of if all you had was a neuromodulatory system to guide you would be a little bit like playing a billion dimensional huddle call. I really loved that analogy that really, it really kind of, it really kind of knocked me off for a loop when, when he said it. And I think it’s really a really profound way to think about, uh, the problem, the computational problem of just having an arousal system. But when I started thinking about that a little bit, um, this actually was, uh, you know, before I’d written the paper that we’ve talked about and originally the paper we talked about actually had a whole other section on the arousal system, but, um, it got, so, um, I don’t know how much swearing I’m allowed to do on your podcast, but it got absolutely destroyed by people because they just, they didn’t like it and whatever I’ll move on, but
Paul 01:31:52 He got fucked by people.
Mac 01:31:55 No, there was actually worse than that. Yeah. Um, it’s, you know, it’s, it’s really in the morning Australia, I’ll try and keep my language. Um, so, um, yeah, so I ended up taking it out and it’s been, it’s formed the basis of multiple other small things that we’ve done on the side. But one of the things that I think could help you solve that multi-dimensional, um, hot or cold problem is if the system that is processing the information, quote unquote, uh, could non-linearly interact with the arousal system, right? So it’s not now, it’s, you’re not now just following the Dover main gradient or following the noradrenaline gradient or the, what you’re saying is if the noradrenaline gradient is high, or if you get a phasic burst of dopamine, whatever it is active at that time, notice it and take advantage of it. Use that window.
Mac 01:32:41 You just had to, to like turn into a purse, firing neurons, to create some long term potentiation. And all of a sudden you’re not playing billion dimensional, uh, hot or cold anymore. What you’re doing is you’re actually playing the game, but, you know, with a little cheat code, which is that if you get even close to something, you can be like, oh wait, forget about everything else. Now I know I’m in the right location. And then you play it again, building dimensional, uh, you know, hot or cold, but only in that one little pocket. And then you get a little bit closer and then you do it again, what’s that that’s grading to set, right? Like that’s kind of, you know, uh, an algorithm that we know works really, really well for a lot of these learning, um, situation. So, um, I’m really eager to see, you know, the much more clever people, uh, than I going into this space and trying to figure out how these systems work together. Um, I think it’s a really interesting time to be a neuroscientist and to try to apply these tools of dynamical systems, but also try to make them make contact with learning. And if anyone wants to talk, you know, hit me up, I’m on Twitter, we can chat,
Paul 01:33:41 Uh, really excited and optimistic. And I feel happy for you because my take is that you feel like you’re in a great position in your career and that the future is wide open. Do I read you correctly there?
Mac 01:33:58 Yeah, I think that’s fair. I I’ve been incredibly fortunate, Paul. Um, you know, I had a really great PhD supervisor who supported me. I then got a fellowship to go over and do research on their work with a world leader in SMRI who then put me in contact with a world leader in computational modeling and then came back to Australia and I’ve got another research on the position where I can spend time to think and read. Um, I live a little bit further away from campus, uh, on the central coast of, of Australia, which has this beautiful, quiet area, lovely beaches, lovely Bush walks and an extremely patient and supportive wife that, uh, you know, allows me to kind of wander around in my head all the day. And so, and, and just a brilliant team around me, great collaborators, lovely, really collegiate and really sort of impassioned young scientists that really pushed me and forced me to kind of stay on my toes the whole time. So I feel incredibly fortunate to, to be in this position. And so my optimism, I think, comes from a just complete lack of understanding of how I’ve found myself in the position that I found myself in, and I want to do as much good work as I can, and I want to be as engaged with the science as I can. Um, but it’s, it’s really just dumb luck to be perfectly
Paul 01:35:12 Well. Okay. So there’s a lot to unpack there, um, because sure luck, I, I think luck is a huge factor and that’s wonderful that you acknowledged that. Uh, but it also, you know, as you said before, we are a gentle, you have to make those decisions, right? So, uh, it’s not dumb luck. It’s more like serendipity and part of your serendipity, I would imagine arises because of your work ethic, but also your range of interests. And what I’m getting at is that the, the nature of your work and its range, uh, it really takes so much effort, uh, to what, what was your phrase wander in your own, um, ignorance wandering? What was the phrase?
Mac 01:35:57 It was something like
Paul 01:35:58 Wondering your own curiosity. Yeah, my ISE, I say ignorance because I’m more of a pessimist, I suppose, but so, so you, you have to like swim in these disparate facts, right. That, uh, in, in the unknown and that, it’s awesome. It’s awesome. But it also takes a lot of work. Right. And focus. So here, so this is what I’m getting at. Uh, how do you confer that ability to other people? Like what did, what the question is? What advice would you give to someone who is interested in trying to think across scales, across temporal scales, across, uh, physical scales, across systems, inter system, uh, with the brain and think holistically, somewhat like you do. Um, you know, not everyone is suited nor, um, would want to do something like this because it’s, it’s nice to just focus on one little brain area and what it might be doing, but do you, do you have advice to your incoming students and to those that you talk to who are kind of wowed by this kind of holistic understanding you possess?
Mac 01:37:04 Uh, I, you can’t see this on the podcast, but I’m blushing, um, as well. Um, I played football for a number of years, um, back when I was younger and could still move without feeling like a creaky old, uh, kind of chair that’s been lifted out in the rain for a few months. Um, and my team was really good and, uh, we won a bunch of championships and we had a lot of fun. We had a really great tight-knit group, but I was never the best player I was always on a good team, but I had to rely on this guy to make this play on this other guy to make that play. And I think that had a huge, a huge effect on me. I think it hit me really hard back in those days that if you want to achieve something, uh, that’s really big and, and requires that, um, coordination, you’ve got to be able to be a part of a team and you’ve got to make that really integral in what you do.
Mac 01:38:02 And I’ve carried that through with me and I’ve tried really hard to, you can maybe talk to my students and collaborators on the side and see if it’s reflecting the feedback. But I consider that to be absolutely crucial is that I’m not doing anything on my own. Everything that I do is about the team that I’m in and the, and the group of scientists that I’m fortunate enough to work with. Um, so that I think is one pot. And I think another thing that comes with that perspective is that you should never ever feel like all the pressure’s on you to do it, everything. Um, you, I think if, if you follow your curiosity, if you are dissatisfied with answers that don’t resonate with what makes sense to you then I think you naturally play that little game of grading descent across the landscape in a really fun way.
Mac 01:38:51 And as, as you said, Paul, it doesn’t always make sense. Um, and there are pots that are really frustrating and there are gaps in the literature that, that you so desperately want filled, but aren’t going to be filled for practical reasons. Um, as a, you know, uh, curiosity driven sort of scientists, uh, in a systems level, those gaps are everywhere. And so I think you have to kind of make a lot of, um, guesses. You have to do a lot of hedging, and I think you have to kind of use the soundboard of your collaborative group and, and yourself really to try to kind of work out where the solid bits of ground are and where the parts that are a little bit more flimsy and you have to be willing to kind of live in that murkiness. There’s a lot of benefits, but, but you know, it’s, it’s not, it’s not the kind of thing where you kind of, um, you want to kind of like launch out on that stuff when you haven’t got a position that’s, you know, relatively solid, you know, I’ve only been out being able to do stuff like this because I would get a, you know, a fellowship based on some, you know, some work.
Mac 01:39:48 And then I would have a few years of stability where I could say, okay, now I’ve got the flexibility to go out and explore a little bit and I can try this and that and the other. And, and, you know, there’s definitely points in this story where, you know, after getting a really bad rejection, uh, you know, you think, man, what did I just waste all this time? Like, what have I done wrong? What have I missed? But then you pick yourself up and you go, okay, I’m going to take the pots to the work. It’s a little bit like, uh, my kids will be watching the Marvel movies and it’s a little bit like at the end of the, all the Ironman ones, but he like takes the thing. He had it, he like throws off like half of the stuff. And then he starts with like the little bit, and then he builds it up.
Mac 01:40:20 I think that’s an underrated pot of trying to like navigate science as well as to like, they call it, you know, don’t be afraid to kill your darlings. You’ve gotta be really willing to kind of depart with things that will useful, but then ultimately don’t work. And I think, again, this is where it comes back to being part of something bigger than yourself being part of a team and not feeling like if your research question ultimately is a dead end, that it’s your fault or something. You’re a scientist to me is someone who’s out there trying to discover something about the world to understand it a little bit better. And it comes in many different forms. It could be done at the patch clamp, trying to work out how that particular cell worked, or it could be up at the broad level asking about ecological interactions in a complex ecosystem, anywhere in between, you can apply scientific thinking. It’s a process, not a set of facts. And so to me, I, you know, I just feel so fortunate to be a part of this process. It’s a lovely job to have, and I feel like a, an absolute moron most days out of the week, uh, based on how much uncertainty I’ve ever everything. But when you catch those little bits of insight, when the puzzle pieces align, when the little idea clicks to you, oh man, it’s, it makes it so worthwhile.
Paul 01:41:29 Well, it really comes through that. Uh, you feel fortunate and excited. Do you know where the phrase kill your darlings comes from?
Mac 01:41:38 I do not actually.
Paul 01:41:39 Oh, it’s Stephen King in reference to heavy handed editing and how, uh, how appreciative, uh, how beneficial that is. So, uh, Matt, thank you so much. Um, it’s been really fun having you finally come on the show. I’m glad we finally got you on and continue the great work, man.
Mac 01:41:55 Thanks, Paul
Paul 01:42:02 Brain inspired is a production of me and you. I don’t do advertisements. You can support the show through Patrion for a trifling amount and get access to the full versions of all the episodes. Plus bonus episodes that focus more on the cultural side, but still have science go to brain inspired.co and find the red Patrion button there to get in touch with me. emailPaul@braininspired.co. The music you hear is by the new year. Find email@example.com. Thank you for your support. See you next time.
0:00 – Intro
6:32 – Background
10:41 – Holistic approach
18:19 – Importance of thalamus
35:19 – Thalamus circuitry
40:30 – Cerebellum
46:15 – Predictive processing
49:32 – Brain as dynamical attractor landscape
56:48 – System 1 and system 2
1:02:38 – How to think about the thalamus
1:06:45 – Causality in complex systems
1:11:09 – Clinical applications
1:15:02 – Ascending arousal system and neuromodulators
1:27:48 – Implications for AI
1:33:40 – Career serendipity
1:35:12 – Advice